Amazon triples GPU orders to secure AI future

Amazon has significantly expanded its procurement of Nvidia graphics processing units, tripling its recent orders to a total of three million chips. This substantial increase builds on earlier plans to integrate over one million GPUs into its cloud infrastructure, with the new hardware set to arrive between 2027 and 2028. According to reports from GN technics/ai (en-US), this move signals a deepening strategic alliance rather than a simple hardware purchase.

The deal includes the latest Nvidia architectures, such as Blackwell Ultra and Rubin. For Amazon, securing this capacity is critical to maintaining its position in the competitive cloud market. However, the long lead time means the financial returns from this investment remain distant, creating a significant gap between the current expenditure and future potential earnings.

Strategic partnership extends beyond hardware

This agreement is not limited to chip sales; it encompasses a broader collaboration on AI factories, networking, and robotics. Nvidia aims to leverage this partnership to meet its revenue targets for its upcoming chip generations. For Amazon, this ensures a stable supply of compute power, which is increasingly scarce in the industry. The scope of the deal reflects how both companies rely on each other to drive their respective AI strategies forward.

The expansion highlights the critical role of infrastructure in the current AI race. By committing to such large volumes, Amazon is positioning itself to handle growing workloads without facing supply bottlenecks. This approach contrasts with the volatility often seen in smaller-scale procurement, offering a level of certainty that is rare in the fast-moving tech sector.

Internal chips reduce Nvidia dependency

Despite the massive Nvidia order, Amazon is simultaneously scaling its own custom silicon business. Its in-house chips, including Trainium accelerators and Graviton processors, have reached a substantial annualized revenue run rate. This dual strategy provides Amazon with negotiating leverage and a hedge against potential supply or pricing issues from Nvidia. It ensures that the company is not wholly dependent on a single external vendor for its core AI capabilities.

The existence of a robust internal chip ecosystem also serves as a long-term safeguard. As the demand for AI computing power fluctuates, Amazon can adjust its mix of internal and external hardware. This flexibility is a key advantage in a market where technology standards and costs can shift rapidly. It allows the company to optimize for performance and cost efficiency over time.

Long lead times pose risks

The primary catch in this deal is the timeline. The newly ordered chips will not be operational until 2027 or 2028. This means the investment is a long-term gamble on the sustained demand for AI services. If market needs shift or technology evolves differently than anticipated, the hardware could become less relevant or overkill. This creates a financial risk that is not immediately visible in current quarterly results.

Investors must weigh the immediate capital expenditure against the distant potential returns. The deal strengthens Amazon’s position as a major buyer, but it also locks in significant costs for years to come. The success of this strategy depends on whether the AI-driven cloud growth materializes as projected. Until the chips arrive and generate revenue, the full impact of this decision remains uncertain.